Red Hat Enterprise Linux Server Cookbook

Chapter 1. Working with KVM Guests

In this chapter, we will cover the following recipes:

Installing and configuring a KVM
Configuring resources
Building VMs
Adding CPUs on the fly
Adding RAM on the fly
Adding disks on the fly
Moving disks to another storage
Moving VMs
Backing up your VM metadata

Introduction

This book will attempt to show you how to deploy RHEL 7 systems without too much of a hassle. As this book is written with automation in mind, I will emphasize on command-line utilities rather than elaborating on its GUI counterparts, which are useless for automation.

This chapter explains how to build and manage KVM guests using the libvirt interface and various tools built around it. It will provide a brief overview on how to set up a KVM on RHEL and manage its resources. The setup provided in this overview is far from the ready enterprise as it doesn't provide any redundancy, which is generally required in enterprises. However, the recipes provided are relevant in enterprise setups as the interface stays the same. Most of the time, you will probably use a management layer (such as RHEV or oVirt), which will make your life easier in managing redundancy.

Note

Libvirt is the API between the user and the various virtualization and container layers that are available, such as KVM, VMware, Hyper-V, and Linux Containers. Check https://libvirt.org/drivers.html for a complete list of supported hypervisors and container solutions.

As most tasks performed need to be automated in the end, I tend not to use any graphical interfaces as these do not allow an easy conversion into script. Hence, you will not find any recipes in this chapter involving a graphical interface. These recipes will primarily focus on virsh, the libvirt management user interface that is used to manage various aspects of your KVM host and guests. While a lot of people rely on the edit option of virsh, it doesn't allow you to edit a guest's configuration in real time. Editing your guest's XML configuration in this way will require you to shut down and boot your guest for the changes to take effect. A reboot of your guest doesn't do the trick as the XML configuration needs to be completely reread by the guest's instance in order for it to apply the changes. Only a fresh boot of the guest will do this.

The virsh interface is also a shell, so by launching virsh without any commands, you will enter the libvirt management shell. A very interesting command is help. This will output all the available commands grouped by keyword. Each command accepts the --help argument to show a detailed list of the possible arguments, and their explanation, which you can use.

Installing and configuring a KVM

This recipe covers the installing of virtualization tools and packages on RHEL 7.

By default, a RHEL 7 system doesn't come with a KVM or libvirt preinstalled. This can be installed in three ways:

Through the graphical setup during the system's setup
Via a kickstart installation
Through a manual installation from the command line

For this recipe, you should know how to install packages using yum, and your system should be configured to have access to the default RHEL 7 repository (refer to Chapter 8, Yum and Repositories, for more information), which is required for the packages that we will use.

Alternatively, you could install packages from the installation media using rpm, but you'll need to figure out the dependencies yourself.

Check the dependencies of an rpm using the following command:

~]# rpm -qpR <rpm file>

This will output a list of binaries, libraries, and files that you need installed prior to installing this package.

Check which package contains these files through this command:

~]# rpm -qlp <rpm package>

As you can imagine, this is a tedious job and can take quite some time as you need to figure out every dependency for every package that you want to install in this way.

Getting ready

To install a KVM, you will require at least 6 GB of free disk space, 2 GB of RAM, and an additional core or thread per guest.

Check whether your CPU supports a virtualization flag (such as SVM or VMX). Some hardware vendors disable this in the BIOS, so you may want to check your BIOS as well. Run the following command:

~]# grep -E 'svm|vmx' /proc/cpuinfo
flags    : ... vmx ...

Alternatively, you can run the following command:

~]# grep -E 'svm|vmx' /proc/cpuinfo
flags    : ... svm ...

Check whether the hardware virtualization modules (such as kvm_intel and kvm) are loaded in the kernel using the following command:

~]# lsmod | grep kvm
kvm_intel             155648  0
kvm                   495616  1 kvm_intel

How to do it…

We'll look at the three ways of installing a KVM onto your system.

Manual installation

This way of installing a KVM is generally done once the base system is installed by some other means. You need to perform the following steps:

Install the software needed to provide an environment to host virtualized guests with the following command:
```
~]# yum -y install qemu-kvm qemu-img libvirt
```
The installation of these packages will include quite a lot of dependencies.
Install additional utilities required to configure libvirt and install virtual machines by running this command:
```
~]# yum -y install virt-install libvirt-python python-virthost libvirt-client
```

By default, the libvirt daemon is marked to autostart on each boot. Check whether it is enabled by executing the following command:

~]# systemctl status libvirtd
libvirtd.service - Virtualization daemon
   Loaded: loaded (/usr/lib/systemd/system/libvirtd.service; enabled)
   Active: inactive
     Docs: man:libvirtd(8)
           http://libvirt.org

If for some reason this is not the case, mark it for autostart by executing the following:
```
~]# systemctl enable libvirtd
```

To manually stop/start/restart the libvirt daemon, this is what you'll need to execute:

~]# systemctl stop libvirtd
~]# systemctl start libvirtd
~]# systemctl restart libvirtd

Kickstart installation

Installing a KVM during kickstart offers you an easy way to automate the installation of KVM instances. Perform the following steps:

Add the following package groups to your kickstarted file in the %packages section:

@virtualization-hypervisor
@virtualization-client
@virtualization-platform
@virtualization-tools

Start the installation of your host with this kickstart file.

Graphical setup during the system's setup

This is probably the least common way of installing a KVM. The only time I used this was during the course of writing this recipe. Here's how you can do this:

Boot from the RHEL 7 Installation media.
Complete all steps besides the Software selection step.
Go to Software Selection to complete the KVM software selection.
Select the Virtualization host radio button in Base Environment, and check the Virtualization Platform checkbox in Add-Ons for Selected Environment:
Finalize the installation.
On the Installation Summary screen, complete any other steps and click on Begin Installation.

Configuring resources

Virtual machines require CPUs, memory, storage, and network access, similar to physical machines. This recipe will show you how to set up a basic KVM environment for easy resource management through libvirt.

A storage pool is a virtual container limited by two factors:

The maximum size allowed by qemu-kvm
The size of the disk on the physical machine

Storage pools may not exceed the size of the disk on the host. The maximum sizes are as follows:

virtio-blk = 2^63 bytes or 8 exabytes (raw files or disk)
EXT4 = ~ 16 TB (using 4 KB block size)
XFS = ~8 exabytes

Getting ready

For this recipe, you will need a volume of at least 2 GB mounted on /vm and access to an NFS server and export.

We'll use NetworkManager to create a bridge, so ensure that you don't disable NetworkManager and have bridge-utils installed.

How to do it…

Let's have a look into managing storage pools and networks.

Creating storage pools

In order to create storage pools, we need to provide the necessary details to the KVM for it to be able to create it. You can do this as follows:

Create a localfs storage pool using virsh on /vm, as follows:

~]# virsh pool-define-as --name localfs-vm --type 
dir --target /vm

Create the target for the storage pool through the following command:
```
~# mkdir -p /nfs/vm
```

Create an NFS storage pool using virsh on NFS server:/export/vm, as follows:

~]# virsh pool-define-as --name nfs-vm --type network --source-host nfsserver --source-path /export/vm –target /nfs/vm

Make the storage pools persistent across reboots through the following commands:
```
~]# virsh pool-autostart localfs-vm
~]# virsh pool-autostart nfs-vm
```

Start the storage pool, as follows:

~]# virsh pool-start localfs-vm
~]# virsh pool-start nfs-vm

Verify that the storage pools are created, started, and persistent across reboots. Run the following for this:

~]# virsh pool-list
 Name                 State      Autostart
-------------------------------------------
 localfs-vm           active     yes
 nfs-vm               active     yes

Querying storage pools

At some point in time, you will need to know how much space you have left in your storage pool.

Get the information of the storage pool by executing the following:

~]# virsh pool-info --pool <pool name>
Name:           nfs-vm
UUID:           some UUID
State:          running
Persistent:     yes
Autostart:      yes
Capacity:       499.99 GiB
Allocation:     307.33 GiB
Available:      192.66 GiB

As you can see, this command easily shows you its disk space allocation and availability.

Tip

Be careful though; if you use a filesystem that supports sparse files, these numbers will most likely be incorrect. You will have to manually calculate the sizes yourself!

To detect whether a file is sparse, run ls -lhs against the file. The -s command will show an additional column (the first), showing the exact space that the file is occupying, as follows:

~]# ls -lhs myfile
121M -rw-------. 1 root root  30G Jun 10 10:27 myfile

Removing storage pools

Sometimes, storage is phased out. So, it needs to be removed from the host.

You have to ensure that no guest is using volumes on the storage pool before proceeding, and you need to remove all the remaining volumes from the storage pool. Here's how to do this:

Remove the storage volume, as follows:

~]# virsh vol-delete --pool <pool name> --vol <volume name>

Stop the storage pool through the following command:
```
~]# virsh pool-destroy --pool <pool name>
```
Delete the storage pool using the following command:
```
~]# virsh pool-delete --pool <pool name>
```

Creating a virtual network

Before creating the virtual networks, we need to build a bridge over our existing network interface. For the sake of convenience, this NIC will be called eth0. Ensure that you record your current network configuration as we'll destroy it and recreate it on the bridge.

Unlike the storage pool, we need to create an XML configuration file to define the networks. There is no command similar to pool-create-as for networks. Perform the following steps:

Create a bridge interface on your network's interface, as follows:

~]# nmcli connection add type bridge autoconnect yes con-name bridge-eth0 ifname bridge-eth0

Remove your NIC's configuration using the following command:
```
~]# nmcli connection delete eth0
```

Configure your bridge, as follows:

~]# nmcli connection modify bridge-eth0 ipv4.addresses <ip address/cidr> ipv4.method manual
~# nmcli connection modify bridge-eth0 ipv4.gateway <gateway ip address>
~]# nmcli connection modify bridge-eth0 ipv4.dns <dns servers>

Finally, add your NIC to the bridge by executing the following:

~]# nmcli connection add type bridge-slave autoconnect yes con-name slave-eth0 ifname eth0 master bridge-eth0

For starters, we'll take a look at how we can create a NATed network similar to the one that is configured by default and called the default:

Create the network XML configuration file, /tmp/net-nat.xml, as follows:

<network>
  <name>NATted</name>
  <forward mode='nat'>
    <nat>
      <port start='1024' end='65535'/>
    </nat>
  </forward>
  <bridge name='virbr0' stp='on' delay='0'/>
  <ip address='192.168.0.1' netmask='255.255.255.0'>
    <dhcp>
      <range start='192.168.0.2' end='192.168.0.254'/>
    </dhcp>
  </ip>
</network>

Define the network in the KVM using the preceding XML configuration file. Execute the following command:
```
~]# virsh net-define /tmp/net-nat.xml
```

Now, let's create a bridged network that can use the network bound to this bridge through the following steps:

Create the network XML configuration file, /tmp/net-bridge-eth0.xml, by running the following:

<network>
    <name>bridge-eth0</name>
    <forward mode="bridge" />
    <bridge name="bridge-eth0" />
</network>

Create the network in the KVM using the preceding file, as follows:
```
~]# virsh net-define /tmp/net-bridge-eth0.xml
```

There's one more type of network that is worth mentioning: the isolated network. This network is only accessible to guests defined in this network as there is no connection to the "real" world.

Create the network XML configuration file, /tmp/net-local.xml, by using the following code:

<network>
  <name>isolated</name>
  <bridge name='virbr1' stp='on' delay='0'/>
  <domain name='isolated'/>
</network>

Create the network in KVM by using the above file:
```
~]# virsh net-define /tmp/net-local.xml
```

Creating networks in this way will register them with the KVM but will not activate them or make them persistent through reboots. So, this is an additional step that you need to perform for each network. Now, perform the following steps:

Make the network persistent across reboots using the following command:
```
~]# virsh net-autostart <network name>
```
Activate the network, as follows:
```
~]# virsh net-start <network name>
```

Verify the existence of the KVM network by executing the following:

~]# virsh net-list --all
 Name                 State      Autostart     Persistent
----------------------------------------------------------
 bridge-eth0          active     yes           yes
 default              inactive   no            yes
 isolated             active     yes           yes
 NATted               active     yes           yes

Removing networks

On some occasions, the networks are phased out; in this case, we need to remove the network from our setup.

Prior to executing this, you need to ensure that no guest is using the network that you want to remove. Perform the following steps to remove the networks:

Stop the network with the following command:

~# virsh net-destroy --network <network name>

Then, delete the network using this command:

~]# virsh net-undefine --network <network name>

How it works…

It's easy to create multiple storage pools using the define-pool-as command, as you can see. Every type of storage pool needs more, or fewer, arguments. In the case of the NFS storage pool, we need to specify the NFS server and export. This is done by specifying--source-host and--source-path respectively.

Creating networks is a bit more complex as it requires you to create a XML configuration file. When you want a network connected transparently to your physical networks, you can only use bridged networks as it is impossible to bind a network straight to your network's interface.

There's more…

The storage backend created in this recipe is not the limit. Libvirt also supports the following backend pools:

Local storage pools

Local storage pools are directly connected to the physical machine. They include local directories, disks, partitions, and LVM volume groups. Local storage pools are not suitable for enterprises as these do not support live migration.

Networked or shared storage pools

Network storage pools include storage shared through standard protocols over a network. This is required when we migrate virtual machines between physical hosts. The supported network storage protocols are Fibre Channel-based LUNs, iSCSI, NFS, GFS2, and SCSI RDMA.

By defining the storage pools and networks in libvirt, you ensure the availability of the resources for your guest. If, for some reason, the resource is unavailable, the KVM will not attempt to start the guests that use these resources.

When checking out the man page for virsh (1), you will find a similar command to net-define, pool-define: net-create, and pool-create (and pool-create-as). The net-create command, similar to pool-create and pool-create-as, creates transient (or temporary) resources, which will be gone when libvirt is restarted. On the other hand, net-define and pool-define (as also pool-define-as) create persistent (or permanent) resources, which will still be there after you restart libvirt.

Building guests

After you install and configure a KVM on the host system, you can create guest operating systems. Every guest is defined by a set of resources and parameters stored in the XML format. When you want to create a new guest, creating such an XML file is quite cumbersome. There are two ways to create a guest:

Using virt-manager
Using virt-install

This recipe will employ the latter as it is perfect for scripting, while virt-manager is a GUI and not very well suited to automate things.

Getting ready

In this recipe, we will cover a generic approach to create a new virtual machine using the bridge-eth0 network bridge and create a virtual disk on the localfs-vm storage pool, which is formatted as QCOW2. The QCOW2 format is a popular virtual disk format as it allows thin provisioning and snapshotting. We will boot the RHEL 7 installation media located on the localfs-iso storage pool (rhel7-install.iso) to start installing a new RHEL 7 system.

How to do it…

Let's create some guests and delete them.

Create a guest

Let's first create a disk for the guest and then create the guest on this disk, as follows:

Create a 10 GB QCOW2 format disk in the localfs-vm pool, as follows:

~]# virsh vol-create-as --pool localfs-vm --name rhel7_guest-vda.qcows2 --format qcows2 –capacity 10G

Create the virtual machine and start it through the following command:

~]# virt-install \
--hvm \
--name rhel7_guest \
–-memory=2048,maxmemory=4096 \
--vcpus=2,maxvcpus=4 \
--os-type linux \
--os-variant rhel7 \
--boot hd,cdrom,network,menu=on \
--controller type=scsi,model=virtio-scsi \
--disk device=cdrom,vol=localfs-iso/rhel7-install.iso,readonly=on,bus=scsi \
--disk device=disk,vol=localfs-vm/rhel7_guest-vda.qcow2,cache=none,bus=scsi \
--network network=bridge-eth0,model=virtio \
--graphics vnc \
--graphics spice \
--noautoconsole \
--memballoon virtio

Deleting a guest

At some point, you'll need to remove the guests. You can do this as follows:

First, ensure that the guest is down by running the following:

~]# virsh list –all
 Id    Name                           State
----------------------------------------------------
-     rhel7_guest                     shut off

If the state is not shut off, you can forcefully shut it down:

~]# virsh destroy --domain <guest name>

List the storage volumes in use by your guest and copy this somewhere:

~]# virsh domblklist <guest name>
Type       Device     Target     Source
------------------------------------------------
file       disk       vda        /vm/rhel7_guest-vda.qcow2
file       cdrom      hda        /iso/rhel7-install.iso

Delete the guest through the following command:
```
~]# virsh undefine --domain <guest name> --storage vda
```
Adding --remove-all-storage to the command will wipe off the data on the storage volumes dedicated to this guest prior to deleting the volume from the pool.

How it works…

The virt-install command supports creating storage volumes (disks) by specifying the pool, size, and format. However, if this storage volume already exists, the application will fail. Depending on the speed of your KVM host disks (local or network) and the size of the guest's disks, the process of creating a new disk may take some time to be completed. By specifying an existing disk with virt-install, you can reuse the disk should you need to reinstall the guest. It would be possible to only create the disk on the first pass and change your command line appropriately after this. However, the fact remains that using virsh vol-create-as gives you more granular control of what you want to do.

We're using the QCOW2 format to contain the guest's disk as it is a popular format when it comes to storing KVM guest disks. This is because it supports thin provisioning and snapshotting.

When creating the guest, we specify both the maxmemory option for memory configuration and the maxvcpus option for vcpus configuration. This will allow us to add CPUs and RAM to the guest while it is running. If we do not assign these, we'll have to shut down the system before being able to change the XML configuration using the following command:

~# virsh edit <hostname>

As you can see, we're using the virtio driver for any hardware (network, disks, or balloon) that supports it as it is native to the KVM and is included in the RHEL 7 kernel.

Note

If, for some reason, your guest OS doesn't support virtio drivers, you should remove the --controller option of the command line and the bus specification from the --disk option.

For more information on virtio support, go to http://wiki.libvirt.org/page/Virtio.

The --memballoon option will ensure that we do not run into problems when we overcommit our memory. When specific guests require more memory, the ballooning driver will ensure that the "idle" guests' memory can be evenly redistributed.

The graphics option will allow you to connect to the guest through the host using either VNC (which is a popular client to control remote computers) or spice (which is the default client for virt-manager). The configuration for both VNC and spice is insecure, though. You can either set this up by specifying a password—by adding password=<password> to each graphics stanza—or by editing the /etc/libvirt/qemu.conf file on the KVM host, which will be applied to all guests.

There's more…

In this recipe, we used "local" install media in the form of an ISO image to install the system. However, it is also possible to install a guest without a CD, DVD, or an ISO image. The --location installation method option allows you to specify a URI that contains your kernel/initrd pair, which is required to start the installation.

Using --location in combination with --extra-args will allow you to specify kernel command-line arguments to pass to the installer. This can be used, for instance, to pass on the location of an Anaconda kickstart file for automated installs and/or specifying your IP configuration during the installer.